Private, affordable, high-performance LLMs via verifiable TEE sharding.
- Shard & run large models across multiple TEEs (Phala) with on-chain attestation of code + model hash.
- Two ways to use:
- Host a shard → profit share + direct usage.
- Buy tokens with ETH on Base → pay-per-query in chat.
- Smart contracts automate payments, usage metering, and revenue split.
Data stays private · Costs are pooled · No single-GPU/TEE limits · Decentralized & fault-tolerant.
TinyLlama split into 2 shards; single /generate endpoint; both TEEs post attestation reports.
- Vertical: more shards → bigger models.
- Horizontal: P2P mesh → failover & scale.
- Demo: https://tee-tee.vercel.app/
- Setup: https://github.com/derek2403/TeeTee/blob/main/TeeDockerFiles/README.md
- Contract (Base Sepolia):
0x396061f4eba244416ca7020fa341f8f6a990d991
Liew Qi Jian · Phen Jing Yuan · Edwina Hon Kai Xin · Cedric Chung Theng Fung · Tan Zhi Wei · Marcus Tan · Lee Wai Yee